Ten stories moved this week, and four of them turn on the same fight. Microsoft's own scientist called his employer's AI training theft. The U.S. Department of Justice calls the same practice fair use. Satya Nadella wants a button a human can press on any AI model. Anthropic wants Congress to write that button into law. Read the rest of today's stories the same way. A pattern holds. Everyone wants to own the rules for AI agents. Few agree on what they should say.
Microsoft Calls Its Own Scraper Theft
A court filing unsealed Oct. 16 names Microsoft's scraper Project Mango1. Microsoft built it to copy webpages for OpenAI's training data1. Brent Hecht, Microsoft's director of applied science, wrote that the copying could be "the largest theft of labor in human history"1. Microsoft told staff to keep the project secret1.
David Buttle runs the publisher coalition Spur. He says the works carry a price that Microsoft and OpenAI skipped. "The defendants' own conduct establishes that the content has a market price," Buttle said. "The case is about their refusal to pay it to these plaintiffs"2.
The U.S. Department of Justice disagrees. Its September brief stated: "Consistent with the national interest in promoting innovation and free expression, the training of AI models on copyrighted material, in and of itself, does not violate copyright laws"3. Opposition briefs from the publishers are due Oct. 2312.
Anthropic's Valuation Talk Drops to $1.5 Trillion
Anthropic told investors its third-quarter revenue hit $16.8 billion4. That is up 45 percent from $11.6 billion in the second quarter4. Revenue started the year at $4.7 billion4. Growth is still real, but it slowed. The second quarter alone grew nearly 150 percent4.
Anthropic's quarterly revenue more than triples to $16.8 billion in two quarters
Anthropic's self-reported quarterly revenue, Q1 to Q3 2026, billions of dollars
Each point is one quarter's revenue Anthropic told investors; the line shows revenue climbing from $4.7 billion to $16.8 billion.
Source: The Information [4]. Chart by The AILately.com Desk.
The numbers behind this chart
| Period | Value |
|---|---|
| Q1 | $4.7B |
| Q2 | $11.6B |
| Q3 | $16.8B |
Some investors now weigh a $1.5 trillion price for Anthropic's November stock sale. That is down from the $2 trillion figure floated weeks earlier5. Reuters first reported the lower number late Thursday, through Investing.com5. Anthropic left both figures unconfirmed5.
Investor talk of Anthropic's IPO price falls 25 percent in one quarter
Valuation figures investors discussed for Anthropic's planned November stock sale
Each column is a valuation investors floated: the first before Anthropic's Q3 report, the second lower figure after it.
Both figures are investor discussions reported before Anthropic or its underwriters set a final price.
Source: Investing.com [5]. Chart by The AILately.com Desk.
The numbers behind this chart
| Item | Value |
|---|---|
| Before Q3 | $2.0T |
| After Q3 | $1.5T |
Nadella Wants a Brake Inside Every Model
Satya Nadella posted his safety plan on X Oct. 106. He wants companies to assume a model could be compromised from the start. Nadella wants a system that lets an authorized person pause or shut down a model mid-task6. "We must assume a model is compromised and contain it from the start," Nadella wrote6. He called the design a "trust architecture." Every meaningful model action should leave tamper-proof records, in Nadella's plan6. Nadella also wants companies to skip betting critical decisions on one model alone. Their systems should open to independent audits too6.
White House Task Force Adopts Its Charter
The White House's Super Intelligence task force adopted its charter Oct. 157. Director of National Intelligence Jay Clayton leads the group7. The charter gives members 120 days to write a risk report7. That report will cover hacking, dangerous biology and model control7. Defense Undersecretary Emil Michael will draft rules for labs to report breaches7. Vice President JD Vance and two Cabinet secretaries sit on the task force too7.
Congress Gets Anthropic's Own Safety Bill
Anthropic pitched a draft AI safety bill to lawmakers this week8. Politico first reported the pitch Oct. 168. The bill lets the government step in only for a true emergency. That means a model risking more than $1 billion in damage, or more than 50 deaths and injuries8. Independent reviewers would register with the Commerce Department8. A reviewer who finds serious risk must tell Commerce within 72 hours8. The bill keeps other federal agencies from pausing AI work on their own8. Only a federal court in Washington could order a wider halt8. Anthropic wrote the thresholds narrow on purpose. A regulator acts only on a true emergency, and skips every smaller complaint8.
OpenAI's Executives Keep Leaving
A dozen senior OpenAI leaders left in 20269. They ranged from the chief revenue officer to the head of robotics hardware9.
Muse Wonder Costs Like Opus, Scores Lower
Meta shipped Muse Wonder Oct. 1210. It costs the same as Claude Opus 5.510. Artificial Analysis scored it six points behind on its intelligence index10.
TSMC Posts Its Best Quarter Yet
TSMC's profit jumped 69 percent to $23.59 billion, its best quarter yet on AI chip demand11.
Six Labs Extract 190 Million Chats From Claude
Anthropic calls it distillation12. That means copying a rival model's answers to train your own12. Six Chinese labs ran nearly 190 million such chats with Claude between May and July12. Alibaba alone ran 151 million of them12.
Muse Gets Shut Out of Checkout
Amazon blocked Muse in September13. Muse hid its identity from shoppers13. The agent held customer logins outside Amazon's approval13. Amazon says a shopping agent should name itself openly. It should work only through a deal the merchant signed13.
By the numbers
- 160,903 works: copied from news publishers by Microsoft's Project Mango, each one skipping a license fee, court filings show1.
- $16.8 billion: Anthropic's third-quarter revenue, up 45 percent from the second quarter's $11.6 billion4.
- $1.5 trillion: the lower IPO valuation some investors now discuss, down from an earlier $2 trillion5.
- $1 billion: the damage threshold in Anthropic's draft bill before the government could block a model8.
- 120 days: the clock on the White House task force's first risk report, started Oct. 157.
- 151 million: Alibaba's share of nearly 190 million Claude chats Anthropic calls distillation12.
- 6 points: how far Meta's Muse Wonder trails Claude Opus 5.5 on Artificial Analysis's intelligence index10.
What to watch
Watch the Southern District of New York for a ruling on summary judgment123. Opposition briefs are due Oct. 23123. Anthropic's investor roadshow will show which valuation sticks once the IPO prices in November5. Congress now holds Anthropic's draft bill, waiting to find a sponsor8. The Super Intelligence task force's 120-day report will show whether Washington leans on Nadella's brake, Anthropic's bill, or a plan of its own678.
Sources
- Annelise Levy, "Microsoft Told Staff to Keep OpenAI Scraping Project 'Secret'," Bloomberg Law, Oct. 16, 2026, https://news.bloomberglaw.com/ip-law/microsoft-told-staff-to-keep-openai-scraping-project-secret
- Jessica Davies and Sara Guaglione, "Paywall violation 'eviscerates' fair use defense: What OpenAI-NYT twist means for publishers," Digiday, Sept. 21, 2026, https://digiday.com/media/paywall-violation-eviscerates-fair-use-defense-what-openai-nyt-twist-means-for-publishers/
- Wendy Davis, "Trump Admin To Court: AI Training Is 'Fair Use'," MediaPost, Sept. 2, 2026, https://www.mediapost.com/publications/article/417630/trump-admin-to-judge-ai-training-is-fair-use.html
- Cory Weinberg, "Anthropic Reports $16.8 Billion in Third-Quarter Revenue," The Information, Oct. 15, 2026, https://www.theinformation.com/briefings/anthropic-reports-16-8-billion-third-quarter-revenue
- Vlad Schepkov, "Some investors push Anthropic to cut IPO valuation to $1.5 trillion - report," Investing.com, Oct. 15, 2026, https://uk.investing.com/news/stock-market-news/some-investors-push-anthropic-to-cut-ipo-valuation-to-15-trillion--report-4913225
- Anthony Ha, "Microsoft's Satya Nadella says AI models need an 'emergency brake'," TechCrunch, Oct. 10, 2026, https://techcrunch.com/2026/10/10/microsofts-satya-nadella-says-ai-models-need-an-emergency-brake/
- Ashley Gold, "Exclusive: White House's AI task force gets to work," Semafor, Oct. 15, 2026, https://www.semafor.com/article/10/15/2026/white-houses-ai-task-force-gets-to-work
- Simon Mugo, "Anthropic proposes AI safety bill with strict intervention thresholds," Investing.com, Oct. 16, 2026, https://www.yahoo.com/news/politics/articles/anthropic-proposes-ai-safety-bill-232641388.html
- Dealroom, "OpenAI's revolving door spins on: 12 executives exit in 2026," Dealroom, Aug. 14, 2026, https://dealroom.co/news/145108-openais-revolving-door-spins-on-12-executives-exit-in-2026/
- Rama Adi, "Meta Muse Wonder review: benchmarks, pricing, and the Full catch," eesel AI, Oct. 13, 2026, https://www.eesel.ai/blog/meta-muse-wonder-review
- Shalu Saraf, "TSMC Q3 Profit Jumps 69% as AI Chip Demand Drives Record Quarter," TipRanks, Oct. 15, 2026, https://www.tipranks.com/news/tsmc-q3-profit-jumps-69-as-ai-chip-demand-drives-record-quarter
- Samuel Nwite, "Anthropic Says Chinese AI Labs Launched 190 Million Claude Distillation Attacks," Tekedia, Sept. 12, 2026, https://www.tekedia.com/anthropic-says-chinese-ai-labs-launched-190-million-claude-distillation-attacks/
- Luis Rijo, "Amazon blocks Meta's Muse AI agent from shopping on its site," PPC Land, Oct. 8, 2026, https://ppc.land/amazon-blocks-metas-muse-ai-agent-from-shopping-on-its-site/